For the traditional method of retrieved land surface temperature (LST) using multichannel thermal infrared information, it is very difficult to improve the accuracy of LST because of the high correlation among the channels, furthermore, the component temperature can not be retrieved. Based on directional law of thermal infrared radiance of non isothermal mixed pixel, the directional model of continuous vegetation has been established and the relationship between the component effective emissivities and variables has been simulated by Monte Carlo method in this paper. Results show the thermal infrared radiance of non isothermal continuous vegetation is the function of the component temperatures T_v, T_s, leaf area index (LAI), leaf angle distribution (LAD), the single leaf emissivity ε_v(θ) and soil surface emissivity ε_s(θ).The variables left to be retrieved are only T_v, T_s, LAI and T_v(0).Theoretically, if we take advantage of four data from different viewing angles, the four variables including the component temperatures would be inversed due to the low correlation among the thermal infrared information of different viewing angles. Except the data from the vertical viewing angle, the other three data used for retrieved component temperatures should be limited to a range of 30 to 50 degrees.